Artificial Intelligence Based Antibiotic Zone Measurement For Disk Diffusion

The disk diffusion tests used in clinical microbiology laboratories to guide antibiotic therapy decisions and monitor antibiotic resistance patterns in bacterial populations, provides valuable information for selecting the most appropriate antibiotic treatment for bacterial infections. Manual measurement of zone diameters in the laboratory with a special ruler or micrometer device is a difficult process that is not sensitive enough and takes a long time. In this study, we proposed a new hybrid model in which computer vision and deep learning-based artificial intelligence techniques are used together in pipelined fashion to measure zone diameters more precisely and rapidly. With our model comprising multiple Convolutional Neural Networks (CNNs) based YOLO architectures for object and text recognition tasks, zone measurement, classification and coloring can be done practically according to European Committee on Antimicrobial Susceptibility Testing (EUCAST) standards over zone images. Our proposed hybrid model achieves on average ${90, 14 \%}, {92, 09 \%}$ and $90,69 \%$ zone, disk and text detection accuracies respectively.